Framing the use of AI in public administration. Algorithmic accountability, Institutional settings  and social trust

Konstantina Kotsiopoulou ¹ and Charalampos Chrysomallidis ²

¹ Adjunct Lecturer and Postdoctoral Researcher, Department of Business and Organization Administration, University of the Peloponnese, Antikalamos, Kalamata 24100, Greece, k.kotsiopoulou@uop.gr (corresponding author)

2 Assistant Professor, Department of Business and Organization Administration, University of the Peloponnese, Antikalamos, Kalamata 24100, Greece, ch.chrysomallidis@uop.gr

Building on our previous work on the institutional embedding of Artificial Intelligence in public administration, this paper argues that technical readiness and use of AI tools in public administration does not contribute to digital transformation of public sector, per se. Beyond that institutional settings and algorithmic accountability constitute necessary preconditions for the legitimate deployment of AI-related systems in government. While previous research demonstrated that AI functions primarily as an operational amplifier, reinforcing existing administrative capacities and weaknesses, the present study shifts attention from adoption to governance.

The paper examines the institutional settings and the mechanisms required to ensure accountability and contestability throughout the lifecycle of algorithmic decision-making. Drawing on the emerging literature on algorithmic governance, public accountability, and AI regulation, it develops a framework structured around three stages: (a) the design and development phase, (b) the operational phase, and (c) the post-decision phase, where citizens may seek explanation, review, or redress.

The central argument is that institutional embedding alone cannot guarantee legitimate algorithmic governance. Even well-developed digital administrations require procedural safeguards that enable oversight, responsibility attribution, transparency, and the effective contestation of automated decisions. The article therefore moves beyond the question of how AI becomes embedded within public institutions and focuses instead on how public administrations can remain legitimate and accountable once algorithmic systems become part of administrative decision-making, in order to ensure and secure social trust.

Using the European AI Act as a regulatory reference point and drawing on recent European experiences, the paper contributes to the literature by linking institutional capacity with accountability mechanisms across the entire lifecycle of algorithmic governance. The analysis is particularly relevant for countries such as Greece, where significant advances in digital government create a policy window for establishing accountability structures before large-scale algorithmic decision-making becomes fully institutionalized.

Keywords: Artificial Intelligence,  Algorithmic Accountability, Public Administration, Institutional Settings, Social Trust

JEL Classification: H83, O33, H11, D73